| 摘要 |
<P>PROBLEM TO BE SOLVED: To improve accuracy of recognition reliability of speech recognition results. <P>SOLUTION: The speech recognition method includes a speech recognition process and a speech document recognition reliability calculation process. The speech recognition process outputs a speech recognition result, in which the reliability of word recognition is added to each word after speech recognition processing is applied to an input speech document. The speech document recognition reliability calculation process calculates context reliability obtained from the relevance of each word to the other words, by reference to sonic reliability obtained from the word recognition reliability and the relevance table showing relevance among words that constitute the speech recognition result, and calculates the speech document recognition reliability, which is integration of the sonic reliability and the context reliability, for each speech document. <P>COPYRIGHT: (C)2012,JPO&INPIT |
